Multi-layer barriers on polymer (PEN) substrate analysis of thin Al2O3 ALD films on fast deposited SiOx buffer layers

Transcription

1 Eindhoven University of Technology BACHELOR Multi-layer barriers on polymer (PEN) substrate analysis of thin Al2O3 ALD films on fast deposited SiOx buffer layers Schalken, J.R.G. Award date: 2012 Link to publication Disclaimer This document contains a student thesis (bachelor's or master's), as authored by a student at Eindhoven University of Technology. Student theses are made available in the TU/e repository upon obtaining the required degree. The grade received is not published on the document as presented in the repository. The required complexity or quality of research of student theses may vary by program, and the required minimum study period may vary in duration. General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. Users may download and print one copy of any publication from the public portal for the purpose of private study or research. You may not further distribute the material or use it for any profit-making activity or commercial gain

2 Multi-layer barriers on polymer (PEN) substrate: Analysis of thin Al 2 O 3 ALD lms on fast deposited SiO x buer layers Jean-Paul Schalken ( ) Under supervision of: Hindrik de Vries and Mariadriana Creatore November 18, 2012

3 1

4 Preface This report is a part of my master's degree program in Applied Physics at Eindhoven University. An internship at a university abroad or at a company is a part of the master's program, in order to apply technical en scientic knowledge in a research environment, other than at our university. I'm very glad that Fujilm gave me the oppurtunity to do an internship. In this way, I can experience the dierences between research in an academic environment and at a company like Fujilm. This can help me with my decisions further in my career. Therefore, I want to thank Hindrik and Adriana for providing me this option. Next to that, I want to thank Hindrik, Sergey and Adriana for their discussions and feedback. Of course I also want to thank Bernadette, Bruno, Marco and Rinie for their technical support, as well as all their other help. At last, I want to thank the people from Eindhoven University which technically supported me, thank you Cristian and Wytze. 2

5 3

6 Contents 1 Introduction: The eld of barrier layers 6 2 Sample preparation Substrate Deposition by a Dielectric Barrier Discharge Atomic layer deposition (ALD) Plasma assisted ALD Analysis methods Water vapor transmission rate (WVTR) Calcium testing Interferometry Spectroscopic Ellipsometry (SE) Results & Discussion Surface analysis of substrates Single layer barrier analysis Sealing a SiO x buer layer with Al 2 O Inuence of longer oxygen plasma exposure Pore analysis via activation energy Conclusions & Recommendations 62 A WVTR Measurement values after aluminum foil experiments 66 Bibliography 69 4

7 5

8 Chapter 1 Introduction: The eld of barrier layers In the challenge to improve electronic devices, steps are made in trying to produce cheaper electronics with better properties. A major cost reduction can be achieved by the roll-toroll production of organic electronics on polymer substrates. Devices which can be produced in a roll-to-roll setup are organic solar cells, organic light-emitting diodes (OLED) as well as inorganic CIGS photovoltaics (thin lms of copper indium gallium diselenide). LED technology is an established technology for lighting, as well as that it is often used as backlight for LCD television screens. Organic LED technology is an emerging technology that allows the production of exible displays.[20] A problem when working with organic material, however, is that devices degrade under the inuence of water vapor.[21] This susceptibility to moisture is induced by the metal cathodes, which get oxidized by water vapor and oxygen, leading to a delamination of the underlying organic layer.[21] Non-emissive black spots are resulting at the surface of the OLED. This oxidation process should be counteracted by the deposition of a barrier layer. Where glass or metal substrates have effective barrier properties themselves, polymer substrates require a barrier for water vapor and oxygen permeation, which should protect the device from degradation.[20] It is estimated that the maximum permeation through a barrier should be around g/m 2 day for water vapor (WVTR) and cm 3 /m 2 day (OTR), resulting in an OLED lifetime of > h.[20] This is the time, that is needed to degrade the reactive cathode of the OLED by water vapor and oxygen. There are several techniques and materials researched to deposit barrier layers. However, only a few of them work as eective barriers for water vapor and for oxygen. Nowadays, lots of research is done in the deposition of barrier layers by atomic layer deposition (ALD). ALD is known for the ability to deposit dense, conformal layers with a low defect density.[9] This is a good technique to deposit barrier layers, especially because the per- 6

9 meation through these kind of barrier layers is mainly driven by defects.[6] The principle of the ALD technique as well as its advantages and disadvantages will be more extensively discussed in sect. (2.3). Since deposition of Al 2 O 3 by ALD is conformal and delivers a dense lm, there is often research performed using these layers as barriers layers. Carcia et al. also did some interesting research depositing Al 2 O 3 barriers, which resulted in an estimated barrier of g/m 2 day at standard conditions of 23 C and 50 % relative humidity (RH).[4] This was derived from experiments at elevated temperatures (38 C and 60 C): via these results, the apparent activation energy ( E act ) could be determined and the WVTR could be extrapolated to room temperature. The measurements were done on a 25 nm, by ALD deposited, barrier layer of Al 2 O 3, deposited on a PEN substrate (polyethylene naphthalate), while the ALD process was performed at 120 C, using trimethylaluminum (TMA) and water vapor as precursors. In their research, they mention the importance of the cleaning process: The cleanliness and surface chemistry of the polymer were critical for achieving thin, defect free lms with ultra low permeability. Therefore they rinsed the samples before loading them into an ALD reactor via a laminar ow hood, operating at class 100 cleanroom conditions.[4] In a later publication, they explained more about their cleaning procedure.[3] Also Keuning et al. deposited Al 2 O 3 barrier layers by ALD. In their ALD process, however, they used an oxygen plasma[12], instead of water vapor to oxidize the Al 2 O 3 lm, opening the possibility to deposit at lower temperatures: 25 C. Due to this low temperature, they used a plasma exposure time of 4 s, to keep the carbon content low. With these conditions, they obtained an intrinsic barrier performance (WVTR) of g/m 2 day for nm barrier layers at 20 C and 50 % RH. Both Carcia et al. and Keuning et al. characterized the barrier performance using a calcium test (Ca test). This test method oers the possibility to measure the intrinsic WVTR in between the defects, excluding water vapor permeation through pinholes (as used in this research project and by Keuning et al.), but a Ca test can also be used to determine the extrinsic WVTR (as used by Carcia et al.). This is shown in g. (1.1). The dierence between the two setups is the epoxy layer, which is used at the setup of Carcia et al. This epoxy layer allows the water vapor, which has permeated through the barrier lm, to diuse all over the dierent calcium spots. This is in contrast to our own setup in which the water vapor, after it has permeated through the barrier, directly oxydaizes the calcium at the specic spot of the pores or pinholes. The Ca test is explained more extensively in sect. (3.2). The deposition rate of ALD, however, is relatively low. Therefore, until recent, it 7

10 Figure 1.1: Setup of the Ca test for the determination of the intrinsic WVTR (left) and the extrinsic WVTR (right, [4] (edited)). The bottom gures are examples of the specic Ca tests. White dots are present in our setup due to the direct contact between the barrier layer and the calcium, while an overal grey area is shown in the results of Carcia et al.[4] was believed that good barrier lms at high deposition rate could only be achieved by a low pressure PECVD process. However, establishing roll-to-roll high volume PECVD deposition of barrier lms in vacuum is not a prerequisite. An alternative to low pressure PECVD is to deposit barrier layers via a dielectric barrier discharge (DBD, sect. (2.2)), since it can easily be implemented in a roll-to-roll production system. Using the so-called atmospheric pressure, glow discharge plasma (APGD plasma), SiO x lms with an extrinsic barrier of < g/m 2 day were demonstrated.[25] Fujilm and TU/e already gained a lot of knowledge on this DBD.[1, 2, 25, 26, 29, 30, 31, 32] A remarkable result in this project was that the barrier layers deposited by the APGD plasma have a WVTR which is strongly dependent on the barrier thickness according WVTR t 3.[8] This typical behavior is shown in g. (1.2). In this gure the WVTR measured by Fujilm is compared to the results of ideal laminate theory (ILT,[27]) Three types of permeation through the SiO x can be distinguished.[27] Permeation can take place via macro-defects (md, size> 1 nm), nano-defects (nd, size= nm) or lattice interstices (la, size< 0.3 nm) as shown in g. (1.3). Since the size of a H 2 O molecule is 0.33 nm, it can be assumed that the permeation through lattice interstices is negligible. 8

11 Figure 1.2: The inuence of the thickness of the barrier layer on the WVTR. The results of research performed at Fujilm are shown and compared to results of ILT. It is shown that some permeation through the lattice is possible, though many orders of magnitude lower than the permeation via nano-defects.[7] Furthermore, it can be said that well deposited barriers are characterized by the fact that there are no macro-defects found, leaving the only transport mechanism to be through nano-scale defects.[6] These nano-defects form columnar molecular size pathways at the edges of grain boundaries, via which H 2 O and O 2 can diuse through a SiO x barrier, to reach the polymer.[6] In principle, the permeation through a barrier can be described by some simple equations from ILT, depending on the permeation through the individual layers P ILT = ( φpolymer + φ ) 1 SiO x using φ polymer = P polymer P tpolymer/t and φ SiOx = tsiox/t (1.1) SiOx And since P SiOx/φ SiOx P polymer/φ polymer, this can be simplied to: P barrier,ilt P SiO x φ SiOx (1.2) From this equation, it can be concluded that the WVTR should be proportional to the inverse of the thickness of the barrier layer (W V T R t 1 ), which is in contrast to the results found by Fujilm, mentioned before (WVTR t 3 ). There can be several reasons for this discrepancy. The most important reason is that ILT assumes a homogeneous uniform barrier. However - as already shown before - this is not the case. The transport 9

12 Figure 1.3: Three types of permeation, as proposed by A.P. Roberts et. al.[27] of water vapor through the barrier is mainly through defects in the material. The average number and size of these defects, depends on the deposition method and on the process conditions, but also on the thickness of the layer, as thicker layers have less defects per unit volume than thinner layers.[27] So the pore structure of the barrier, as well as the surface roughness can have a large inuence on the barrier performance. Furthermore, water vapor will interact with the SiO x surface, forming SiOH groups at dangling Si O and Si bonds.[6] The H 2 O molecules can get adsorbed at the surface of the nano-sized pores, which requires energy to again remove them, as well as that these adsorbed molecules tend to hinder the permeation through these pores, simply because the pores get thinner. However, this eect of adsorbing water vapor at the edges of the pores is said to be less important for SiO x, compared to AlO x or SiN x, because the binding energy for SiO x is much lower.[8] At last, it can be possible that the SiO x barrier is damaged by the permeation of the water vapor, e.g. cracks can be created or enlarged.[6] Furthermore, from eq. (1.1) it is also possible to derive a multi-layer model for permeation through a barrier ((1.3),[3]). This is important when producing a barrier consisting of multiple layers. When using multiple layers a mismatch between the various pores of these layers can result in a much lower permeability of the barrier, since the permeating gas is forced to (partly) diuse through the intrinsic material, instead of through the pores.[33] For a single layer barrier, diusion through pores would be the main driving force for permeation. A barrier of intrinsic material - where the permeant should diuse via lattice interstices - is much better, resulting in a lower permeation. 10

13 1/P barrier = 1 /P polymer + 1 /P layer1 + 1 /P layer2 (1.3) At last it, it is shown that the type of permeation (via macro-defects, nano-defects or lattice interstices) can be related to the activation energy of the barrier ( E act ). For a specic permeant, the activation energy increases if the barrier is deposited with less defects.[33] For oxygen permeation, E act can vary from ±30 kj /mol for a weak barrier with lots of large defects (same as substrate), up to ±100 kj /mol for ideal SiO x where permeation is only possible via lattice interstices.[33] For water vapor permeation on the other hand, E act varies from kj /mol for poor barriers (same as PET substrate) up to ±84 kj /mol for ideal, silica glass barriers.[6, 27] This activation energy relates the permeation ( Π in [ mol /m 2 s atm]) to the temperature via:[33] Π = Π 0 exp ( Eact /RT) (1.4) in which Π 0 is a constant and R the ideal gas constant. From this the activation energy can be determined: E act = R ( 1 /T) ln (Π /Π 0 ) (1.5) Analysis of the pores structure are important in this project. Since in this project analysis are done to increase the deposition rate of SiO x barriers while sustaining the barrier quality; the WVTR should remain the same, or even decrease. In general, an increase of the deposition rate results in worse SiO x barrier lms. The barrier structure becomes more porous, which enhance moisture permeation. Since water vapor permeation is mainly through interconnected nano-pores, the idea is to deposit a thin lm on top of this SiO x layer, which is able to seal this percolated structure. Atomic layer deposition (ALD, sec. (2.3)) is used for this, since it is a conformal and defect free deposition technique. When depositing a thin layer of Al 2 O 3 ( nm) by ALD on top of the SiO x layer - deposited by DBD - a good barrier may be deposited, while the deposition process can be performed much faster. This research focuses on the option to sustain low moisture permeation through HDR SiO x layer by sealing this buer layer with a thin Al 2 O 3 layer. This is also depicted in g. (1.2). For example the minimum thickness of the Al 2 O 3 layer has been investigated to maintain a good barrier quality. Furthermore, the interaction between the layers deposited by CVD and ALD is analyzed; are the layers intermixing and in what kind of way? Also the inuence of the amount of carbon in the layer is researched. If the Al 2 O 3 layer is deposited 11

14 Figure 1.4: Typical behavior of the SiO x layer deposited by APGD-CVD. The aim of this project is to improve the high deposition rate deposited SiO x to the same level of the red bullet. This is done by sealing the HDR SiO x layer with Al 2 O 3, deposited by ALD. with higher or lower carbon content, what can be the inuence of this on the WVTR. The quality of the barrier is also analyzed, by the determination of the activation energy. This activation energy can be used to determine the WVTR at atmospheric conditions, but also gives information about the pores structure of the barrier. 12

15 13

16 Chapter 2 Sample preparation The samples for this research project are prepared on exible substrates (PEN or PET, see sec. (2.1)). A SiO x lm ( 70 nm) is deposited on these substrates using chemical vapor deposition via an atmospheric pressure glow discharge (APGD-CVD), which is also called dielectric barrier discharge (DBD, sec. (2.2)). Subsequently the SiO x layer is sealed by a thin layer of Al 2 O 3 ( nm) using plasma assisted atomic layer deposition (PA-ALD, sec. (2.4)). The APGD-CVD and ALD deposition techniques, as well as background on the used substrates is provided in this chapter. Furthermore, standard thermal ALD is explained (see sec. (2.3)), since this is the basic principle of PA-ALD. 2.1 Substrate The idea is that the barriers are protecting products like exible electronics against deterioration by water vapor. Therefore, there are several requirements for the substrate. The most obvious properties are exibility, transparency and robustness, but next to that, it is also desired that the substrate has a low coecient of thermal expansion (CTE) amongst others, so that it can be easily processable at dierent temperatures. A high CTE may promote cracks in the barrier material, which would drastically increase the water vapor transmission rate (WVTR). There are several substrates meeting these requirements. Research in this kind of eld is often performed on PEN, poly(ethylene 2,6-naphthalate).[3, 25] The structure formula for PEN is shown in g. (2.1). This organic polymer, from DuPont Teijin Films, is heat stabilized and it has an anti-blocking layer on the back. The anti-blocking layer is desired when winding up the polymer, taking care that the two sides of the polymer do not stick to each other, creating cracks already before deposition. The total thickness of the PEN substrate is 100 µm. 14

17 Figure 2.1: Structure formula of PEN [poly(ethylene 2,6-naphthalate)] Figure 2.2: Structure formula of PET [poly(ethylene teraphthalate)] Another optional material, which can be used as a substrate for the analysis of the barriers is PET, poly(ethylene teraphthalate). This is even more known, because of the worldwide use in packaging or drinking bottles. It is cheaper than PEN, but the properties are not as good as for PEN. PET is also be delivered with an anti-blocking layer on the back. PEN and PET can also be delivered with a transparent planarizing layer (a TP layer). With the TP layer the surface becomes smoother and has fewer defects. This PEN/TP (or PET/TP) surface is protected by a so called inter leaf foil (ILF), in order to protect the substrate when winding it up or unwinding it later on. The ILF is removed just before the deposition starts, keeping the substrate surface as clean as possible. In that way, the ILF reduces the probability of particle contamination during the web handling, otherwise more defects are generated. Defects are the major permeation mechanism in the permeation of water vapor or oxygen through the barrier[27], so it is important to prevent the surface from all kinds of particles, which could create defects.[4] Since the TP layer is a product of Fujilm and it is not exactly known, the surface properties of this organic planarizing layer are investigated using interferometry. These results are compared to the bare PEN and bare PET layers (results shown in sec. (4.1)). This is done by interferometry. The working of interferometry will be explained in sec. (3.3). 2.2 Deposition by a Dielectric Barrier Discharge When industrially producing barriers for exible electronics, it is desired to combine the production of coatings for large area substrates together with reasonable low production 15

18 Figure 2.3: a) Standard lamentary discharge, b) Uniform diusive atmospheric pressure glow discharge.[32] costs. In this view, a fast depositing setup is preferred. To keep the production costs low, a considerable volume of uniform non-thermal plasma is needed, together with an atmospheric pressure setup. The atmospheric pressure setup oers large cost reduction, because vacuum equipment is not needed anymore. Next to that, when working with at atmospheric pressure, an in-line, roll-to-roll production becomes possible, which can be combined with organic processing steps already performed at Fujilm. In this way, the production should be relatively cheap and it could run at high deposition rates and web speeds. A good option to perform these kind of depositions is by a dielectric barrier discharge (DBD).[17] A problem with atmospheric pressure DBD, however, is that the plasma discharge usually splits in several tiny laments. Though it is possible to create a uniform, diuse mode DBD, using specic gas mixture compositions, a specic power and a specic operation frequency. Furthermore, matching and stabilization electronics are needed. This uniform, diusive DBD mode is called APGD (atmospheric pressure glow discharge).[32] The dierence between the standard lamentary discharge and an APGD is shown in g. (2.3). The DBD system consists of two rotating metal drums placed ±0.5 mm above each other. These drums are heated up to 80 ºC by circulating oil through it. By applying a voltage to the coils, the drums work as two electrodes of a capacitor. The power for these electrodes is delivered by a SEREN LF2001 generator with a frequency of 200 khz. The power can be varied, but is in the range of W, depending on which type of layer is deposited. This is explained further in this section. Furthermore, the power is pulsed. This is important in avoiding the accumulation of negative ions in the discharge region. 16

19 Figure 2.4: Roll-to-roll setup of APGD-CVD. On the left side the polymer substrate is unwinded, while on the right it is winded up again. Also the gas injection system is shown on the left. The accumulation of negative ions leads to the formation of dust, increasing the pinhole density in the lm.[32] The setup can also be seen in g. (2.4). This gure shows two identical webs in between the electrodes. Both webs work as a dielectric, reducing the conduction current between the electrodes so that the probability of an arc discharge is reduced. Furthermore, it can be seen that this is a roll-to-roll setup. On the left the two coils are unwinding, while the coils on the right are winding up the foil. However, to prevent the samples from winding induced damage, the samples are cut from the web before any mechanical contact. The plasma itself is generated between the electrodes by a glow discharge. The electrodes are placed ±0.5 mm from each other so that a gas mixture can ow in between the electrodes. That gas is injected by a gas injection system, which is placed just on the left of these electrodes. As mentioned in the introduction, the idea is to deposit a SiO x layer with a high deposition rate (HDR). The resulting SiO x layer is a percolated layer (columnar structures of interconnecting defects) with a relatively high WVTR, but which can be sealed by atomic layer deposition (ALD). The gas ows for this type of layer are set at: 15 slm of N 2, 1.0 slm of Ar, 1.8 slm of O 2 and 5.5 g/hr of TEOS. TEOS (tetra-ethyl orthosilicate) is the precursor, reacting with the oxygen, so that SiO x is deposited. Under these conditions, 17

20 the thickness of the deposited layers is approximately 70 nm. The growth speed for this HDR SiO x is: v growth ( HDR SiO x ) = 24.6 nm/s (2.1) On the other hand (as shown in g. (1.4), together with the previously explained HDR deposited SiO x ), is it also possible to deposit a low deposition rate SiO x barrier. When using the correct parameters, a much better barrier can be deposited, but - as said - this is deposited much slower, resulting in a growth speed as shown in eq. (2.2). Furthermore, this barrier does not have such an improvement when depositing an ALD layer on top of it. This will be extensively discussed in sect. (4.3). This LDR barrier is deposited using gas ows of 15 slm of N 2, 1.0 slm of Ar, 1.0 slm of O 2 and just 1.0 g /hr of TEOS. The power will here be 600 W, while for the HDR barrier, it will be 650 W. v growth ( LDR SiO x ) = 4.90 nm/s (2.2) Especially the oxygen in the mixtures is important for the plasma chemistry. Of course for depositing the silicon oxide, promoting the precursor decomposition, but also in the carbon content of the lm. For low oxygen concentration, lots of carbon can be found (up to 60 %), while for higher oxygen concentrations, the carbon content can be reduced to ± 5 %.[32] Furthermore, to get a better idea of the discharge deposition process, the plasma is simulated using BOLSIG+. This program is used to determine transport coecients and rate coecients, which are calculated from collision cross sections. The exact formulas used to do this are shown in Hagelaar and Pitchford[11]. It should be mentioned that this is more like a rough estimation in order to get some idea of the plasma properties. The determination of the exact properties is behind the scope of this research, which is more focused on the properties of the deposited lms, than that it really focuses on the techniques that are used. As mentioned, the cross sections for momentum transfer (elastic or momentum, see [11]), excitation, ionization and detachment are imported into the BOLSIG+ simulation program. The cross sections of gasses which are used - Ar, N 2 and O 2 - are already in the database of the program. Only the cross sections of TEOS should be added. These are determined by Morgen et al.[22] and they are shown in g. (2.5). The ionization cross sections of the four materials are also shown in g. (2.6). In order to correctly simulate the plasma, some parameters should be determined. At 18

21 Figure 2.5: TEOS cross sections by Morgan et al.[22] These are the cross sections for elastic momentum transfer (Q e ), inelastic momentum transfer (Q m ), excitation (Q v, for two dierent excitation energies), ionization (Q i ) and detachment (Q d ). Figure 2.6: Ionization cross sections (m 2 ) of all four materials in the gas mixture, which are ionized by the glow discharge, as a function of energy (ev). 19

22 Table 2.1: Resulting data of the plasma simulation analysis. ow fraction rate coecient ionization rate [ m3 /s] [ m3 /s] TEOS 5.5 g /hr N slm Ar 1.0 slm O slm rst, the electric eld, E, can be approximated by a parallel plate capacitor[37]: E = V d = 2kV 0.5mm = V/m (2.3) Together with the density, n, determined from the ideal gas law: n = N V = p RT = E/n can be determined in [Td] via: 10 5 Pa J /K mole 353K = 34.1 mole /m 3 (2.4) E n = 4 106V/m 34.1 mole /m 3 = V m 2 /mole = 196 Td (2.5) Next to this electric eld value, of course the composition of the gas mixtures should be known. Using a molar mass M T EOS = g /mole, the fractions of the various gasses in the mixture are shown in table (2.1). The resulting rate coecients and the ionization rate per component in the plasma are also shown in this table. It can be concluded that oxygen and nitrogen are the most important components in the plasma. Furthermore, lots of other data can be generated, however, the most important might be the mean energy of the plasma, being 5.23 ev. 2.3 Atomic layer deposition (ALD) If the permeation through a barrier is controlled by pinholes, a multi-layer barrier may help to increase the barrier performance and reduce the water vapor transmission rate (WVTR).[15] Although Fujilm is able to make decent barriers of SiO x with DBD, the production process for these barriers is relatively slow. As shown in g. (1.4), it is desired to have a faster deposition process, while the barrier quality remains similar. When trying to establish this kind of barrier, SiO x is deposited at high deposition rate (HDR) by DBD 20

23 and covered by a very thin layer ( nm) of Al 2 O 3. The deposition of the Al 2 O 3 layer can be performed by atomic layer deposition (ALD). This can also be deposited relatively fast - because the Al 2 O 3 layer is very thin - so that the total production process can be performed fast enough and the production of barrier layers becomes feasible both in production time and in costs. In this section the basic principle of ALD will be explained, while in the next section, remarks are made about the dierences between plasma assisted ALD (PA-ALD) - the deposition method which is used in this research project - and standard thermal ALD, as well as that the specic conditions and deposition parameters are explained. The basic principle of the most common form of ALD - thermal ALD - is a deposition process which consist of four steps, which are consistently performed when the previous step is nished.[9] In the rst of these steps, the surface is exposed to an aluminum depositing precursor, being trimethylaluminum (TMA, Al(CH 3 ) 3 ). Second, the ALD deposition chamber is cleaned from previously used gasses by ushing it with a nonreactive gas. This is called purging and in this case, argon is used for this. The purging step should be included, because TMA and the second precursor, H 2 O, are heavily reactive when brought together. After this purging step, the second precursor - water vapor - is lead into the chamber. This is an oxidizing precursor, depositing oxygen atoms on the surface, so that the desired AlO x layer is created, as well as it prepares the surface for a new dose of TMA. At last, the chamber is again purged with argon, after which the next cycle can start by TMA exposure again. The actual Al 2 O 3 lm is grown in the second and fourth step of the ALD cycle by exposing the surface to the precursors, TMA and H 2 O. The precursors react with specic surface groups ( OH groups and CH 3 groups respectively) so that TMA deposits aluminum on the surface and the water vapor deposits oxygen, leaving specic surface groups behind again for the next step to take place. The actual chemical surface reactions of this deposition process are shown in eq. (2.6) and (2.7). The molecules which are attached to the surface are marked with an (*)[9]: AlOH*+Al(CH 3 ) 3 AlOAl(CH 3 ) 2 *+CH 4 (2.6) AlCH 3 *+H 2 O AlOH*+CH 4 (2.7) The reactions are graphically shown in g. (2.7). In the rst half of the ALD cycle (eq. (2.6)), the TMA adsorbs at a hydroxide group at the surface, after which CH 4 is formed from the hydrogen atom at the surface and one of the methyl groups of the 21

24 TMA. Immediately after the formation of this CH 4 group, the CH 4 is reabsorbed from the surface, leaving aluminum at the surface, together with two methyl groups. So in the rst half of the ALD cycle, the hydrogen atom at the surface is replaced by aluminum, carrying two CH 3 groups. This is also shown in the second (middle) illustration in g. (2.7). For the second growing step, after purging the reactor with Ar, the surface is exposed to water vapor. The water adsorbs at the Al atom at the surface. Then again CH 4 is formed from one hydrogen atom of the water molecule and a methyl group, which was deposited at the surface during the rst step. Almost immediately after formation, the CH 4 is again desorbed from the surface, leaving a hydroxide group at the surface. In this way an oxygen atom is deposited at the surface, while the surface is ready to interact with TMA again.[35] This is also shown in bottom part of g. (2.7). In principle, when performing both half-cycles together, each containing one of the deposition steps and a purging step, a full mono-layer of Al 2 O 3 can be deposited. Although it is shown that it takes more than one cycle to deposit a complete mono-layer (explained later on in this section), a general equation can be derived for the thermal Al 2 O 3 ALD by correctly adding up eq. (2.6) and eq. (2.7): 2Al(CH 3 ) 3 + 3H 2 O Al 2 O 3 + 6CH 4 (2.8) On the left hand side of this equation, both precursors are present: TMA and water, while on the right the exhausted methane shows up, together with the aluminum oxide, which is deposited on the surface. This can also be concluded from g. (2.7). On the left site of this gure, the precursors and the exhausted methane are also shown, together with the amounts of all these substances. The deposited Al 2 O 3 is of course shown in the layer itself (on the right side of the gure). This kind of thermal Al 2 O 3 ALD can be performed because of the high binding energy of the Al O bond.[9] Forming this bond brings the system in a lower, preferable energy state, since it is exothermic. Therefore during the rst step (eq. (2.6)), the aluminum atom of the TMA prefers sticking to the surface oxide, causing the formation and desorption of a CH 4 molecule. The same holds for the second step (eq. (2.7)), where the oxygen atom prefers creating a Al O bond, which again results in the desorption of a CH 4 molecule. Fig. (2.8) shows the dierence in energy between the various steps of the ALD process. On the left side the various energies of the rst step are shown (the adsorption of TMA), while on the right side the specic energies of the reaction of the changed surface with water are shown. In the latter, the normal numbers are the reactions of Al(OH) 2 CH 3 with H 2 O, while the numbers in parentheses are the values of the reactions of Al(OH)(CH 3 ) CH 3 22

25 Figure 2.7: Principle of thermal Al 2 O 3 ALD using TMA and H 2 O as (metal- and oxide- ) precursors. The upper gure (on the right) shows a OH terminated surface at the beginning of an ALD cycle. After purging with Ar, the rst precursor (TMA) is lead into the reactor. The TMA adsorbs at the O surface atom via the Al atom. Immediately after that, a CH 4 molecule is formed and desorbed from the surface. The Al atom with the two remaining CH 3 groups takes the place of the H atom, as shown in the middle gure. Now the reactor is again purged with Ar to remove the released methane and the remaining TMA. Then water vapor is lead into the reaction chamber, so that OH groups can replace the methyl groups, while producing CH 4 again. The resulting surface is shown in the bottom gure (on the right); it is the same, with the dierence that an extra layer of Al 2 O 3 was deposited on the surface. 23

26 Figure 2.8: Energetics of thermal Al 2 O 3 ALD. The reaction can take place because of the strong Al O bond. On the left side the surface energy is shown while adsorbing a TMA molecule and, after that, desorbing methane. On the right side energies are shown for the adsorption of H 2 O for both surface molecules of Al(OH) 2 CH 3 (normal text) and Al(OH)(CH 3 ) CH 3 (parenthetical).[35] with H 2 O. Both are about equally present in the Al 2 O 3 layer. There are various reasons to use ALD as a technique to deposit the Al 2 O 3 on top of the SiO x layer. The main reasons for using an ALD deposition technique are that ALD is conformal - covering the whole area, also for really thin deposition layers; it is (almost) defect free - which results in very good electrical properties - and the surface deposited by ALD is smooth.[19] These good properties are established, because the ALD process is a self limiting, which means that the amount of deposited material is determined by the amount of surface adsorption sites initially available rather than by the particle ux impinging on the surface.[18] So for the rst half of the ALD cycle (the TMA exposure step), if there is sucient TMA near the surface the deposition is not dependent on the amount of TMA, but on the number of OH groups on the surface, which are able to adsorb the TMA (see gure (2.9)). The same holds for the second half of the cycle. Both steps are limited by the number available surface sites and not by the amount precursor, assuming that there is enough precursor available to cover the entire surface. In principle, when using ALD it should be possible to deposit one mono-layer per ALD cycle. Unfortunately, this is not the case: on average Å/cycle is deposited[36, 9], while one mono-layer of Al 2 O 3 is estimated to be 3.8 Å.[9] This dierence can be explained by two dierent reasons. First, it can be possible that an apparent available surface 24

27 Figure 2.9: In this gure, it can be seen that if enough TMA is available near the surface, the growth per cycle is constant.[14] site is not available. An example of this is steric repulsion; if there is a TMA molecule adsorbed at a surface site, it can be dicult for the adjacent surface site to also adsorb a TMA molecule. The methyl groups which are left at the surface are then screening the available surface site. Problems with the removal of the reaction byproducts (methane) can also counteract adsorption at all available surface surface sites. The second problem limiting the growth is the desorption of adsorbed precursor material. Especially for higher temperatures (e.g. 300 ºC) an adsorbed molecule can again be desorbed due the high available thermal energy.[35] Together with problems like decomposition or condensation of the precursor or incomplete reactions due to the lack of energy, these problems are shown in g. (2.10). This gure also shows that there is only a relatively small temperature window for which ALD is possible. 2.4 Plasma assisted ALD As mentioned before, not standard thermal ALD is used, but plasma assisted ALD. Although plasma assisted ALD (PA-ALD) and thermal ALD are quite similar, PA-ALD has some advantages over thermal ALD. The main advantage is that PA-ALD can be performed at much lower temperatures; even at room temperature. This is of course interesting from a production kind of view. The main dierence is the use of an oxygen plasma, instead of water vapor, as an oxidizing precursor. Due to the use of this plasma, reaction mechanisms also change. The 25

28 Figure 2.10: ALD window. There is a limited range op temperatures at which ALD can be performed.[9] rst half of the ALD cycle (eq. (2.6)) remains the same - using TMA as a precursor and only exhausting CH 4 - but the second part becomes a more combustion type of reaction. That reaction, however, is not as clearly dened as with thermal ALD. The combustion of the CH 3 groups should result in combustion products, like CO 2 and H 2 O, but due to the presence the oxygen plasma, CO 2 can be decomposed, resulting in CO.[12] The decomposition possibilities are also shown in the equations below[14]: e + CO 2 e + CO + O (2.9) e + H 2 O e + H + OH (2.10) Due to these reactions, various reaction products are found. This is also shown in g. (2.11): As mentioned before H 2 O and CO 2 will be found during the plasma exposure step in the ALD cycle, but also CO (due to decomposition of CO 2 ), OH (immediately adsorbed at the surface, but formed due to decomposition of H 2 O) and C x H y (formed out of the reaction of some CH 3 groups (desorbed from the surface due to the presence of the oxygen plasma) and H (left from the decomposition of H 2 O)) are detected when analyzing the various gasses.[12] The precursors for these reactions can be brought into the deposition chamber of the Oxford Instruments FlexAl in two ways. After pumping the chamber down to a pressure of 10 6 bar, the rst option is to simply lead the precursors directly into the chamber, so that the reactions can take place. If the one of the reactions is nished, the remaining precursor, as well as the reaction products should be pumped o and the reaction chamber should be purged. In order to clean the reaction chamber well enough, purging should 26

29 Figure 2.11: Reaction products after plasma exposure in PA-ALD.[14] Figure 2.12: Scheme representing the timing of the various gas ows in the ALD cycle. It is also shown when power is applied to the RF coil, ionizing the oxygen, so that a plasma is created. be done for a relatively long time. Therefore the second option is to use a noble, non reactive, carrier gas (Ar) to lead the precursor to the reaction chamber. In this way, the ALD reaction remains the same, but purging times can be decreased. In g. (2.12) a timing scheme is shown of which gasses are lead into the reaction chamber. The gasses are continuously pumped o. This can be seen in g. (2.13), which shows the reaction chamber setup. The TMA is lead into the reaction chamber from the side, while the oxygen is lead into the chamber from above, passing a radio frequent coil (RF), which can ionize the oxygen to enhance the oxidation step in the ALD cycle. This setup is called a remote plasma, meaning that the plasma is created further above the sample surface, but there is a small, low energy, particle ux going to the surface.[16] At last, of course the growth speed for this setup is important, since the project is focusing on the increase of the deposition speed in order to improve the feasibility of the production of barrier layers. ALD is, however, a relatively slow process, but because of 27

30 Figure 2.13: Setup of the ALD reaction chamber. its properties only a very thin layer of Al 2 O 3 is needed. Using the cycle as shown in g. (2.12) a growth speed can be determined, in order to compare ALD with deposition by DBD. This is of course important for the production of these barriers. v growth = nm /s (2.11) It can be concluded that using this setup, thin lms of Al 2 O 3 can be deposited, sealing the SiO x layer. PA-ALD is chosen to be the method to deposit the lms. It is preferred to use PA-ALD instead of thermal ALD, because the lms can be deposited at lower temperatures, while remaining high lm quality, such as conformality and a low defect density.[9] Furthermore, growth rates are higher and the purging steps can be shortened[12], resulting in a decreased deposition time, attractive for industrial production. 28

31 29

32 Chapter 3 Analysis methods 3.1 Water vapor transmission rate (WVTR) The water vapor transmission rate (WVTR) of the barriers is determined by a Deltaperm Permeation Tester by Technolox. This apparatus determines the extrinsic WVTR and is particularly sensitive for low WVTR barriers. By measuring the vapor pressure of the water vapor which permeates through the barrier, the WVTR can be determined accurately for barriers with WVTR down to ± g/m 2 day. The Deltaperm works with a dierence in pressure between two chambers, which are separated by the placement of a 50 cm 2 sample. A measurement can be started after both of these chambers are pumped down to 0.01 Torr (top part of g. (3.1)) and after the testing conditions are created in the upstream chamber. The testing conditions are created by letting water vapor in the upstream chamber, up to the right humidity. The 90 % RH (at 40 ºC) corresponds to an upstream pressure of Torr. Due to the pressure dierence, the water vapor will diuse through the barrier (middle part of g. (3.1)), ending up in the downstream chamber. In the downstream chamber, the pressure is continuously measured. Due to this water vapor ow, the downstream pressure will increase from 0.01 Torr to maximum 1.0 Torr (bottom part of g. (3.1)). If the downstream pressure reached 1.0 Torr, the downstream chamber is pumped down. Furthermore, if the upstream pressure is decreased due to permeation through the barrier, extra water vapor is let in by opening a water vapor valve for 150 ms. This is checked every minute. At last, the whole system is placed in an oven, which can be set at the desired temperature (40 ºC). Then by taking the derivative of the downstream pressure (which is directly related to the total amount of water vapor which has permeated through the barrier), the WVTR can be determined in [ g /m 2 day]. After some start-up eects, the pressure change will be continuous. 30

33 Figure 3.1: The working of the Deltaperm.[23] The barriers should be used at atmospheric conditions (e.g. 20 ºC and 50 % relative humidity (RH)). Experiments, however, are often performed at other conditions in temperature and relative humidity, e.g. 38 ºC / 85 % RH[4], 20 ºC / 50 % RH[13, 15] or 23 ºC / 100 % RH[5]. Also our experiments are not performed at atmospheric conditions, but at 40 ºC and 90 % RH. When using tougher conditions, measurements can be performed faster, because the permeation is higher. That is because the permeation (P) depends on both the solubility (S) of water vapor into the barrier and the diusion (D) through the barrier.[27] This is shown in g. (3.2) and in eq. (3.1). P = S D (3.1) At last, it should be mentioned that there are some measurements performed on 100 µm aluminum foil. Since this foil can be assumed to be a very good barrier for the water vapor, this can be used to determine the minimum detection limit of this apparatus. This minimum detection limit may be controlled by (air) leakage between the two chamber or via the inlet of water vapor from the outside into the downstream chamber. This can permeate via the grease and rubber, which are used to place the sample properly. The average of these measurements (ve measurements are performed) is (3.6 ± 0.5) 31

34 Figure 3.2: Principle of a permeation experiment. The permeation is both inuenced by the amount of water vapor and the diusion through the barrier.[23] 10 4 g/m 2 day. This value can be subtracted from the WVTR values measured during the experiments. Therefore in Appendix (A), the results are shown after subtraction of the average of this aluminum foil test. 3.2 Calcium testing Next to the extrinsic water vapor transmission rate (WVTR) determined by Deltaperm, the intrinsic WVTR can be determined using a so called calcium test. Using this method the WVTR is determined in the regions in between pinholes (or pores of interconnecting defects), while the permeation through a barrier, measured using an extrinsic method like Deltaperm, is generally considered to be governed by pinholes (LDR SiO x ) or pores of interconnecting defects (HDR SiO x ).[27, 8] This intrinsic WVTR method shows if the barrier itself is good enough to reach the target permeation of g/m 2 day. The calcium test is based on the change in light transmittance of calcium after its reaction with water vapor: Ca + H 2 O CaO + H 2 (3.2) Ca + 2H 2 O Ca(OH) 2 + H 2 (3.3) Since Ca itself is not transparent, while CaO and Ca(OH) 2 are, the amount of water - which has permeated through the barrier - can be determined by the change in light transmittance. From these results, together with a reference measurement, the permeation itself can be determined using equations (3.4) and (3.5).[3] The conditions for these measurements are 20 ºC and 50 % relative humidity (RH). 32

35 Figure 3.3: The various layers in a Ca test. The substrate, together with the barrier is covered by several small squares of Ca. The whole is encapsulated by SiN x : H so that permeation of water vapor only goes via the barrier. Due to the permeation of water vapor through the barrier, the Ca becomes transparent, because of the oxidation to CaO and Ca(OH) 2. W V T R = ρ (CaO) D (CaO) t m (H 2O) m (CaO) (3.4) W V T R = ρ (Ca (OH) 2 ) D (Ca (OH) 2 ) t 2m (H 2O) m (Ca (OH) 2 ) (3.5) It is shown that it is both possible to form CaO and Ca(OH) 2 (eq. (3.2) and (3.3)). Therefore the WVTR can also has to be determined for both of these (eq. (3.4) and (3.5)) and they should be used in their correct ratio. However, when working out eq. (3.4) and (3.5), it can be determined that for the same change in light transmittance, D, the dierence between these two options is less than 3 %.[3] For this kind of WVTR determination, 40 nm of Ca is deposited by thermal evaporation on top of the barrier. The Ca is prepared in two arrays of 3 3 squares of Ca, each with a size of 5 5mm. The Ca is encapsulated by a 300 nm thick layer of PE-CVD deposited SiN x : H. According to Philips, the SiN x : H layer has a barrier of 10 7, so that the change in transmittance is mainly due to permeation through our barrier. This is also shown in g. (3.3). The setup is calibrated by measuring the [Ca] /[O] ratio by Rutherford backscattering (RBS) and the transparency at the same time. 3.3 Interferometry Another technique which is used to analyze the samples is interferometry. A Veeco Wyko NT9100 is used as an optical proler, which is able to determine surface height with an 33

36 Figure 3.4: Resulting gure of an interferometry scan. At each position the height of the surface is determined. accuracy of 1 Å. When analyzing the surface structure, the surface roughness and the number of defects can be determined. This is done using a beam splitter, which splits a laser beam, sending half of the signal to the sample, while the other half is send to a mirror. Depending on the phase dierence between both reected waves, the distance to the surface can be determined. This can be done very accurately. From these results a surface height prole can be obtained. To correct for surface tilting and curving of the substrate an algorithm to lter the low spatial frequencies was applied. This algorithm is available in the software package. Results are shown in g. (3.4). The colors indicate the surface height as shown in the index. In this project interferometry is used to determine the surface roughness and the number of surface defects of the various possible substrates and deposited samples. In order to determine the surface roughness, a scan of the surface is done using a magnication of 100x. The eld of view is µm, which is digitized with a resolution of ( pixels). From the results of the surface height at each spot a 2D plot can be determined, using colors indicating the surface height. An example of such a gure is shown in g. (3.4). Such a gure can, therefore, be used to determine the surface roughness. In fact, there are various numbers related to surface roughness. The most general one, which is applicable here, is the R a value. To determine R a, rst the average surface height, z, is determined (see eq. (3.6)). Then the absolute deviation from this average is determined for each point; this is averaged again (see eq. (3.7)). 34

37 z = 1 A zda (3.6) R a = 1 A z z da (3.7) Furthermore, the number of defects can be determined using interferometry. This is important since defects can lead to pinholes, which largely decrease the performance of the barrier layer. Pinholes, originating from the connection of defects and pores, increase the WVTR enormously by allowing water vapor to ow unhindered through a path of interconnecting defects and pores in the barrier. The defect analysis is done by the determination spots with a certain area, which stand out of the (average) surface a certain height. The last one is the threshold value and is varied, while the area has been kept constant. The threshold value can be seen as a measure for the size of the defect, while the area, is just for excluding very small features and measurement errors (spikes). So for example, an area of nm should completely stand out of the surface for 5 nm. The number of defects (or other features) can be set out in a graph as a function of the threshold value, in order to compare various surfaces. A single measurement is done using a magnication of 40x over an area of mm 2, while 12 measurements are done per sample. At last, there is also a special method from Fujilm to determine the number of pinholes. In this method, the pinholes in the barrier layer can be visualized and can be detected by interferometry. Using a magnication of 5x, an area is scanned of 5 5 mm. Per sample, ve measurements were done at ve dierent spots. Statistical analysis can be done on these measurements to determine properties of these pinholes, however, the most important property is the average number of pinholes per cm Spectroscopic Ellipsometry (SE) The thickness of the various layers of the samples is determined using spectroscopic ellipsometry (SE). Using this technique, the thickness of the SiO x and the Al 2 O 3 layers can be determined, as well as possibly present intermix layers due to the possible penetration of Al 2 O 3 into the SiO x. These measurements are performed using a Woollam M2000 Multi-Angle Ellipsometer. The principle of SE is shown in g. (3.5). A linearly polarized light beam is send to the sample. The polarization of the light is changed as a function of the wavelength due to the 35

38 Figure 3.5: The basic principle of SE.[28] electrical and structural properties of the sample layers. The change in the polarization can be depicted using psi (Y) and delta (D). These are the variables measured as a function of the wavelength of the light. These can be converted to the complex reection coecients r p and r s, parallel (p) and perpendicular (s) to the plane of incidence, via ρ tan Ψ exp (i ) r p r s Erp /Eip E rs/eis (3.8) An optical model should be used to t the results found for Ψ and. All refractive indexes and absorption coecients come together in ρ: tan Ψ exp (i ) = ρ (N 0, N 1, N 2, d, θ 0 ) (3.9) In this equation, only the refractive indexes and absorption coecients of a model with a single thin lm on a substrate are included, as shown in g. (3.6) (left). The model which is used in our research is also shown in (3.6) (right). In g. (3.6) it can also be seen that the barrier consists of several layers. The model consists of the PEN substrate ( 100 µm), together with the SiO x and Al 2 O 3 barrier layers. At the backside, there is a thin anti-blocking layer (sect. (2.1)), however, to make sure that there is back reection, the backside is scratched. Most of this anti-blocking layer is also remove by this. At last, it should also be mentioned that the samples are anisotropic. Therefore it depends in which direction of the sample, the measurements are performed. As is shown by Roelofs[28] the refractive index of PEN is larger in web roll direction, than in the direction perpendicular to that. It can be assumed that this dierence in refractive index is caused by the tension on the web during the (un-)winding process. Because of the anisotropy it is important that each time a SE measurement is performed, the sample is oriented the same way. It is standard that the measurements are performed in the direction perpendicular to the web roll direction. 36

39 Figure 3.6: On the left a basic illustration of SE is shown. The incoming light ray is reected at the various interfaces, where the refractive index changes. Also part of the light is transmitted. On the right the SE model for the deposited barriers is shown, containing the substrate (PEN, together with an anti-blocking layer on the backside) and layers of SiO x and Al 2 O 3. The backside of the samples is scratched, so that only reection of the deposited layers is seen. 37

40 Chapter 4 Results & Discussion In this chapter the experimental results are presented and discussed. The results are discussed starting with the available substrates (sect. (4.1)), followed by single layer barriers (sect. (4.2)) and subsequently the sealing of HDR SiO x with a thin layer of Al 2 O 3, both with a standard ALD process (sect. (4.3)) and a longer oxygen plasma exposure time in the ALD cycle (sect. (4.4)). The last section of this chapter (sect. (4.5)) is dedicated to the determination of the activation energy of the barrier layers prepared in sect. (4.4). For each section individually, the results are discussed and explanations for the various phenomena are proposed after the presentation of the bare results. The explanations are supported by literature. Everything is brought together in a nal conclusion in chapter Surface analysis of substrates It is explained in sect. (2.1) that for this project four dierent substrates are available. Because of the various requirements of the substrates - like transmittance of visible light, mechanical exibility, thermal stability, smoothness of the surface and of course that it should relatively cheap - there are just a few possible substrates, mainly polymers. PEN (polyethylene naphthalate) and PET (polyethylene terephthalate) are examples of substrates which meet the requirements. Both substrates can also be provided with a planarizing lm to make the surface even smoother, this is called a TP-layer. Both PEN and PET, as well as their potential TP-layer are already more extensively described in sect. (2.1). Two characteristics of the dierent possible substrates are investigated using interferometry. In order to compare PEN, PEN/TP, PET and PET/TP the surface roughness and the number of surface defects are determined using interferometry. The specications in determining these properties are stepwise explained in sect. (3.3). 38

41 Figure 4.1: View of the surfaces of PEN, PEN/TP, PET and PET/TP. It can already be seen that the PEN/TP is much smoother than the others. The gures shown are measured using a magnication of 100x and have an area of µm each. Height scales are the same for all four gures and are shown in the middle (ranging from 5 nm to +10 nm). First, the surface roughness of the various substrates is determined. A 3-dimensional view of the substrate surfaces is shown in g. (4.1) and the results of the surface roughness measurements are shown in g. (4.2) respectively. For the 3D view, one of the 18 measurements at dierent spots is depicted (magnication of 100x, scanning area of µm). The average surface roughness for PEN, PET and PET/TP was found to be 0.66 ± 0.05 nm, 0.64 ± 0.10 nm and 0.76 ± 0.07 nm respectively, while the average surface roughness of the PEN/TP was lower, 0.56 ± 0.06 nm. Next to the surface roughness, the number of surface features (defects) is analyzed. This is important because they potentially lead to macro-defects in the barrier lm. The analysis of the number of surface features is also done by interferometry, but using a 40x magnication, which results in a broader scanning area of µm. As shown in g. 39

42 Figure 4.2: Results of the surface roughness analysis of the dierent substrates. The surface roughness of PEN/TP is signicantly lower than the roughness of the other substrates. The data are averaged over 18 measurements at dierent spots (magnication of 100x, scanning area of µm). (4.1) the surface contains several features, which are considered to be a result of defects near the surface. The number of features can be determined using a minimum spot size and a threshold value. First the average surface height is determined, which is corrected for the fact that the surface (and the stage) is tilted and that the surface might be bend. Then, only the larger features are taken into account, using a threshold value for the height of the defect. This value is varied in order to determine the size (especially the height) of the defects. At last, the features should have an area of nine pixels (625 nm 2 ) being higher than the threshold value. This is to ensure that no measurement errors intermix with the feature counting. This is a relatively large area, also due to the low lateral resolution of interferometer ( 1 1 µm) compared to the height resolution (up to 1 Å). Due to this low lateral resolution it might be possible that large thin features (spikes) are spread out to broader features with a much smaller peak height. Therefore, the size of the features might be a bit of an under estimate, depending structure of the features near the surface. Using this procedure, the number of defects on the surface is determined, as well as an estimation is given for the size of these defects via the threshold value. This is shown in g. (4.3). It can be seen that independent of the threshold value, there are less defects on the PEN/TP substrate than on the other substrates. Finally, it should be mentioned that, although the density of features is large, the total measured area is relatively small. Therefore the total number of measured features for threshold values higher than 7 15 nm 40

43 Figure 4.3: Analysis of the number of surface defects. It is shown that the number of surface defects is much lower for the PEN/TP than for the other substrates: bare PEN, PET/TP and bare PET. The data are averaged over 12 measurements, with each an area of mm 2. is, depending on which substrate is analyzed, actually zero. From g. (4.1), (4.2) and (4.3) it can be concluded that the surface roughness of the PEN/TP surface is lower than the surface roughness of the others (R a = 0.56 nm compared to 0.66, 0.64 or 0.76 nm for PEN, PET and PET/TP respectively), as well as that there are less defects found on the PEN/TP substrate. Although there are much more requirements on the substrate - like exibility and a low coecient of thermal expansion - at least from these results, PEN/TP looks favorable as a substrate material, because a smooth substrate with few, small defects, leads to smoothly deposited barrier layers, with less defects and a lower water vapor permeation, which is better for protecting exible organic electronics for oxidation.[4] However, it should be mentioned that the porosity of the substrate also plays a role in the barrier performance.[36] This will be discussed in sect Furthermore it is interesting to see that for PET (0.64 ± 0.10 nm), the substrate itself is smoother than that of PET/TP (0.76±0.07 nm), while the TP-layer should increase the smoothness of the substrate. This can be explained by the overview of the substrate in g. (4.1); it can be seen that PET/TP has a bit of a bumpy surface structure, while bare PET has some big features. Because the surface roughness is determined by the deviation from the average surface height (eq. (3.6) and eq. (3.7)), the PET/TP has lots of contributions 41

44 from the bumpy structure, while PET only has a few from its spike-like features. The measured surface roughness of PET can therefore be lower than that of PET/TP, but the working as a substrate for depositing a barrier might be better for PET/TP, because smooth barrier growth on this substrate can be easier, since the features are more attened out. Since PEN and PEN/TP are performing better (especially PEN/TP) research is continued with these substrates to deposit barriers on for this barrier research. Next to that, PEN substrates are also included to determine the inuence of the TP-layer on the barrier performance. In the next section, PEN and PEN/TP are compared to determine which of those is a better substrate material for the deposition of single layer barriers. Another reason to choose for PEN and PEN/TP is the knowledge of Fujilm from previous research, that these substrates are outperforming PET and PET/TP as a substrate material. 4.2 Single layer barrier analysis In the previous section, the available polymer substrates are analyzed on their surface roughness and the number of surface defects. In this section it will be researched what the barrier properties will be for a single barrier layer, deposited on the previously analyzed substrates. It is important to analyze these single barriers to determine whether the ideal laminate theory holds for our multi-layer barriers. Our barrier will consist of a SiO x buer layer and sealing layer of Al 2 O 3. As shown in sect. (2.2) and (2.4), the SiO x will be deposited by atmospheric pressure, glow discharge, chemical vapor deposition (APGD- CVD) and the Al 2 O 3 will be deposited by plasma assisted atomic layer deposition (PA- ALD). For the SiO x it is possible to quickly deposit a buer layer (HDR SiO x ), which does not have a real barrier itself, but which can be used together with a thin Al 2 O 3 layer as a barrier. However, when slowly depositing a SiO x layer (LDR SiO x ) it is also possible to deposit a moisture barrier, which has barrier in the order of 10 3 g/m 2 day. In this section, three types of single barrier layers on polymer substrate are analyzed. As mentioned in the previous section, PEN and PEN/TP will be used as substrates, while Al 2 O 3, LDR SiO x and HDR SiO x will be analyzed as a single barrier layers in this section. To start with, the water vapor transmission rate (WVTR) of Al 2 O 3 on PEN and PEN/TP is analyzed. The analysis is performed by depositing layers of varying thickness on top of the PEN and PEN/TP. This thickness series is created by changing the number of ALD cycles. In order to determine which Al 2 O 3 layer thickness is desired to obtain a good quality barrier, a series is deposited using 15, 38, 150 and 375 ALD cycles. The 42

45 Figure 4.4: Results of the Deltaperm WVTR test of Al 2 O 3 on both PEN (diamonds) and PEN/TP (squares). The test is performed at 40 ºC and 90 % RH. WVTR of these samples is determined using Deltaperm (sect. (3.1)) and the results are shown in g. (4.4). The layer thickness, obtained by varying the number of ALD cycles, can be estimated by the growth speed of PA-ALD deposited Al 2 O 3 on silicon wafer, being 0.13 nm /cycle. Using this growth speed, the thickness of the deposited layers will be around 2, 5, 20 and 50 nm respectively. However, due to nucleation eects - as explained by Wilson et al. [36] - especially for thin Al 2 O 3 layers, the actual thickness can be less. Therefore, measurements are performed using spectroscopic ellipsometry (SE, sect. (3.4)). These are performed, using the really basic model of Al 2 O 3 on PEN; so without any extra intermixing or planarizing layers. The results of the layer thickness analysis are shown in g. (4.5). If we start by analyzing the numerical values of the WVTR results, then it can be concluded that a barrier can be deposited with a W V T R = g/m 2 day, using 150 cycles of Al 2 O 3 PA-ALD. This is in close agreement to Groner et al. who found a barrier of g/m 2 day for 26 nm of Al 2 O 3, deposited by thermal ALD and measured using an extrinsic HTO test.[10] However, this project focuses on an increase in deposition rate, while sustaining or increasing the barrier quality. The deposition rate for this barrier is relatively slow (eq. (2.11)). Although a relatively thin layer is needed, compared to the SiO x layers we come to later, the deposition time for a layer with a W V T R = 43

46 Figure 4.5: Results of SE measurements. For Al 2 O 3 on PEN the thickness is shown as a function of the number of ALD cycles. A linear t is performed to determine the growth rate of the Al 2 O 3 PA-ALD on PEN g/m 2 day using PA-ALD is 678 s. In order to decrease this deposition time, one might use spatial ALD[24], in which the TMA exposure and oxygen exposure steps are not separated in time (by a long purging step) but in spatial position (by areas of inert gas). However, ALD remains a relatively slow process. A second interesting conclusion which can be drawn from g. (4.4) is that PEN is performing better as a substrate than PEN/TP. Although there is not much known about the TP-layer, it is shown in the previous section that it is smoother and has less surface features than the bare PEN substrate. The only way this signicant deviation can be explained is by the porosity of this TP-layer. Polymer substrates are porous from itself, resulting in Al 2 O 3 nucleation sites inside the surface of the polymer, instead of on top of the polymer. Therefore much more ALD cycles are needed in order to end up with a smooth Al 2 O 3 layer. This can also be seen in g. (4.6) and is more extensively reported and explained by Wilson et al.[36] So if the TP-layer is more porous than the bare PEN substrate, Al 2 O 3 nucleation sites will diuse deeper into the substrate, which results in the fact that it takes more ALD cycles to reach nice layer-by-layer growth and that for the same amount of ALD cycles the number of defects will be larger, reducing the barrier quality and increasing the WVTR. A third interesting point in g. (4.4) is the increase of the WVTR for the Al 2 O 3 layers with an estimated thickness of 50 nm, which were deposited using 375 ALD cycles. 44

47 Figure 4.6: Nucleation and growth during Al 2 O 3 ALD on polymer.[36] This increase should be seen together with the pinholes analysis of the samples deposited on PEN in g. (4.7). This gure looks very similar to the gure on the WVTR. For the thickest sample, an increase of the WVTR is related to the number of pinholes in the sample, which are originated by stress in the material (stress results in small cracks). Therefore it can be concluded that the permeation through the barrier is mainly driven by permeation through pinholes. This is also conrmed by the observations of Erlat et al.[6] and the models of Roberts et al.[27]. As already mentioned in the introduction and the chapter on the sample preparation (sect. (2.2)), it is also possible to deposit a barrier using only SiO x on a substrate. The key is that depositing this kind of barriers is quite a slow process. Although it is much faster than ALD, the deposition rate should still be increased. The barrier performance of this LDR SiO x is also determined as an average over two measurements: W V T R = g/m 2 day. Using the deposition rate as shown in eq. 2.2, the deposition time of this barrier can be determined: 24 s for a 118 nm barrier. Next to the LDR SiO x, also the WVTR of the HDR SiO x is determined. This is done as a reference for the measurements which combine this fast deposited SiO x layer together with a thin layer of Al 2 O 3. The HDR SiO x itself, however, has not a real barrier function; the WVTR is determined to be g/m 2 day for a layer thickness of 74 nm. This thickness determined using spectroscopic ellipsometry (SE, (3.4)) as shown in g. (4.1). 45

48 Table 4.1: SE results of single layer barriers. Figures are shown for each type of single layer barrier, including Ψ and as measured and as generated by the t. The thickness, the refractive index and the WVTR (as determined via Deltaperm measurements) are also presented, together with the MSE-value of the t. The lower the MSE-value, the better the t. 46

49 Figure 4.7: Pinholes analysis of Al 2 O 3 ALD on PEN. 4.3 Sealing a SiO x buer layer with Al 2 O 3 To increase the deposition rate, while aiming at a barrier with W V T R = 10 3 g/m 2 day, both techniques of depositing barrier layers are brought together. On a PEN substrate, a 80 nm HDR SiO x layer is deposited. This is a porous layer with low barrier quality, which can be deposited much faster than the LDR SiO x. The dierence in deposition rate between these two types of SiO x is a factor 8. However, the problem is that this HDR SiO x is not a good barrier itself (W V T R = g/m 2 day). Therefore the HDR SiO x is sealed by a thin layer of Al 2 O 3, deposited by PA-ALD (sect. (2.4)). Both the ALD step and the APGD-CVD step can be performed much faster in this way. The ALD step can be faster, because less cycles are needed and the APGD-CVD step can be performed faster, because the barrier performance requirements for this layer are not as high anymore. Again a series is deposited in which the thickness of the Al 2 O 3 layer is changed, ranging from 0.5 to 20 nm. Since it is assumed that very thin layers are already able to provide a decent barrier quality, the number of ALD cycles is chosen to be: 4, 7, 11, 15, 38, 75 and 150. To determine the barrier performance of this bi-layer barrier, the WVTR is again measured using Deltaperm. The results of these measurements are shown in g. (4.8). In the top of this gure the reference WVTR is given, now for both bare PEN and PEN / HDR SiO x. Furthermore, the thickness of the layers is determined using SE. This is shown in g. (4.9). But again, the determination of the layer thickness can deviate from the expected 47

50 Figure 4.8: Results of the WVTR test of a Al 2 O 3 sealing layer on HDR SiO x. PEN is used as a substrate. The test is performed at 40 ºC and 90 % RH thickness, estimated by the growth rate of Al 2 O 3 ALD. This deviation is strongest for the thinnest of the Al 2 O 3 layers and can be caused by nucleation inside the pores of the SiO x, but also has to do with the way the results are t. Since the refractive indexes of the SiO x and Al 2 O 3 layers are close to each other (see table (4.1)), they are dicult to separate. Therefore the thickness of the reference sample of PEN / HDR SiO x is used as the thickness of the other samples, so that the thickness of the SiO x layer is xed and not tted when the thickness of the Al 2 O 3 is determined. The thickness of the Al 2 O 3 layer, however, can also be estimated by the growth speed of Al 2 O 3 on silicon wafer, because the growth speed is more or less the same on all kinds of substrates.[36] Using a growth speed of 0.13 nm /cycle, the thickness of the sealing Al 2 O 3 will be around 0.5, 1, 1.5, 2, 5, 10 and 20 nm, respectively. The results of the Deltaperm water vapor permeation test show that a decent barrier can already be deposited using 15 cycles of PA-ALD ( 2 nm). When depositing such a thin sealing layer (W V T R = g/m 2 day) it is possible to produce a decent barrier much faster, as well as that it has a barrier performance, which is a factor 2 3 better than that of a single LDR SiO x layer (W V T R g/m 2 day). Next to the Deltaperm measurements, the samples are analyzed by a calcium test (Ca test, sect. (3.2)). For this Ca test, the WVTR is determined at 20 ºC / 50 % RH by measuring the change in the transmittance of visible light of the calcium deposited on 48

51 Figure 4.9: SE results of the thickness of the Al 2 O 3 sealing layer on HDR SiO x. PEN was used as a substrate. the barriers. An example of how the samples look in a Ca test is shown in g. (4.10). The spots on the calcium layer are due to permeation of water vapor through pores of interconnecting defects in the barrier. The WVTR is measured at places in between these spots (intrinsic). The measured WVTR is shown in g. (4.11) and the results of the analysis on the defects are shown in g. (4.12). From the WVTR results of the Ca test it can be concluded that a layer thickness of 5 nm of Al 2 O 3 (= 38 ALD cycles) delivers the best result. This still results in a factor > 2 better barrier performance than for the sample of 15 ALD cycles. This is a larger improvement than for intrinsic WVTR of the Deltaperm, but it can be seen in the same range, due to the larger errors in the Ca test, e.g. for the sample which had 15 ALD cycles W V T R = 1.5 ± g/m 2 day, while for sample with 38 ALD cycles it is measured that W V T R = 4.9 ± g/m 2 day. Another point of interest is that the WVTR of SiN x encapsulation layer itself is close to the results the barrier layers. Therefore it is reasonable that the calcium on top of the barriers is also oxidized by permeation through the encapsulation layer. The WVTR of the barriers might be lower than the WVTR values shown in g. (4.11). A subtraction of the reference measurements - which can give a rough estimate of the barrier performance itself - could then lead to a WVTR of the barrier of g/m 2 day. As was mentioned in the introduction this is the target quality of a barrier, which can be used as a barrier for exible electronics. This target value, however, is an extrinsic value of the WVTR of a barrier. 49

52 Figure 4.10: Example of how the samples look at the start (left two), half-way (middle) and at the end of a Ca test (right). Figure 4.11: The intrinsic WVTR as a function of the number of Al 2 O 3 ALD cycles, as determined by a Ca test. 50

53 Figure 4.12: The number of pinholes (structures of interconnecting nano-defects) as a function of the number of Al 2 O 3 ALD cycles, as determined by a Ca test. So by comparing the results of the Ca test to the target WVTR of g/m 2 day, it is assumed that the intrinsic quality of the barrier can be deposited over larger areas, without defects or pinholes. This can be a hard task, but by careful cleaning of the samples and deposition in ultra clean environments this might be possible. This is shown by Carcia et al. but also by internal research at Fujilm.[4, 8] When taking a look at the number of defects, this number drops signicantly when sealing the HDR SiO x layer with only 1.5 nm of Al 2 O 3 (= 11 ALD cycles) and becomes lower than the SiN x encapsulation layer. This behavior was also seen at the Deltaperm measurements, although it was not as clear there, the WVTR already started to drop signicantly. The number of defects might even be lower, because of particle contamination (dust) during the sample transport. The sealing could than be counteracted. A last important point in g. (4.11) and g. (4.12) is about the sample on which Al 2 O 3 was deposited during 11 ALD cycles (W V T R = g/m 2 day). This sample is produced after all other measurements on the extrinsic WVTR were performed, since it appeared to be in a critical region where the WVTR drops drastically within the range of a few extra ALD cycles of Al 2 O 3. For the extrinsic WVTR this is no problem, since these measurements are performed relatively quick after the deposition of the sample. For the Ca test, however, all measurements are performed at the same time, while this specic sample much later. (The time between the deposition and the actual start of the Ca test was 45 days for sample with 11 ALD cycles, while for the other samples that time was 51

54 73 days.) Therefore it is possible that this fresher sample performs better in a Ca test than the others, resulting in a lower WVTR value in this test. This single result is an indication that the barrier is degrading rapidly in time; even though they were stored carefully in special membrane boxes. The fact that the Al 2 O 3 part of a barrier can be degrading rapidly is also shown by Dameron et al.[5] The degradation of Al 2 O 3 will also be discussed in sect. (4.5). Since the ALD step remains relatively slow compared to the deposition rates of the SiO x in the roll-to-roll system and since the deposition rate is limited by the slowest of the consecutive steps, it is also possible to deposit the ALD layer on top of LDR SiO x. This is also done using 2 and 5 nm of Al 2 O 3 on top of the 70 nm LDR SiO x. The resultant barriers had a WVTR of g/m 2 day and g/m 2 day respectively. These results - being more or less the same - are again signicantly better than the results of Al 2 O 3 on HDR SiO x. However, the deposition rates of both layers are too low. So the barrier performance is largely improved by 2 5 nm Al 2 O 3 lm, deposited by a conformal, defect-free technique, such as PA-ALD. For the HDR SiO x this improvement is about a factor of 350, while for LDR SiO x the improvement factor is much smaller, being a factor 7. However, this is still a signicant improvement, since it is shown in g. (4.4) that a 2 nm Al 2 O 3 layer has no barrier working itself. The improvement cannot be explained by the ideal laminate theory (eq. (1.3)), so there has to be some kind of interaction between the layer. The large increase in barrier performance can be explained by the synergy between the porous SiO x layer and the Al 2 O 3 layer, which is conformally deposited by the ALD. It is determined in earlier research that a SiO x barrier comprises of a percolated structure and the permeation through this barrier is mainly through columnar structures of interconnecting nano-defects.[6, 33, 8] Therefore it can be concluded that an improvement of this barrier - by a thin layer of Al 2 O 3 - should be assigned to the inuence of this thin layer of Al 2 O 3 on the pores of the SiO x layer: hence, the pores are sealed so that permeation through these pores of interconnecting nano-defects is blocked. This sealing principle can be conrmed by the dierence in permeation results between the HDR SiO x and the LDR SiO x layers together with their improvement when sealing it with a thin Al 2 O 3 layer. The fast deposited SiO x barriers have a worse barrier performance, due to a higher pore density.[27] The dierence in improvement can then be seen as the possibility to improve the barrier. The HDR SiO x barrier has a porous structure with lots of interconnecting nano-pores, resulting in a bad barrier performance. But due to the high pore density, lots of these pores are sealed by the Al 2 O 3, which results results 52

55 Figure 4.13: Illustration of the sealing of pores by ALD. The porous structure of the SiO x layer (left) is rst lled by Al 2 O 3 (middle). When depositing more layers conformal layer-by-layer growth will smoothen the surface. The improvement, however, will not be as large anymore. in a factor 350 improvement in barrier performance. On the other hand, for the LDR SiO x layers, there are much less pores - which already results in a decent barrier performance from the beginning - leaving less pores available for the sealing step of the Al 2 O 3. So in the case of LDR SiO x, there are less pores present to be sealed, resulting in a smaller improvement in barrier performance. Next to that, it is assumed that the defects in LDR SiO x barrier layers are much larger, which makes them harder to seal with a thin layer of Al 2 O 3.[8] When looking back to g. (4.6), the existence of a critical sealing layer thickness can be explained. Since nucleation of the Al 2 O 3 layer starts inside the pores of the SiO x layer. It then takes a few ALD cycles before layer by layer growth will start. The pores are really sealed when the growth process has reached layer by layer growth. As shown in g. (4.13), the growth process starts by nucleation inside the pores. After that the pores are sealed by a process which consequently lowers the pore diameter. After a critical number of ALD cycles, the pore diameter has become so small that H 2 O cannot permeate through it. It appears that most pores have a diameter < 5 nm, since after 15 cycles of ALD sealing, the WVTR has dropped drastically. This pore size is conrmed by Erlat et al.[6] 4.4 Inuence of longer oxygen plasma exposure In order to further lower the WVTR and to obtain a better insight in the sealing and nucleation processes, the oxygen plasma exposure time in the ALD cycle is increased up to 10 s (instead of 2 s). When using a longer oxygen plasma exposure time, the carbon content in the ALD lm will decrease, as shown in research done at Eindhoven University. This is proven by varying the oxygen plasma exposure time and the determination of the carbon content via RBS (Rutherford Back-scattering spectrometry) measurements. By lowering the carbon content in the lm, the structure of the Al 2 O 3 will be more crystalline and it would contain less defects (lattice defects). Therefore the WVTR can be lower by 53

56 enlarging the oxygen plasma exposure time. At Eindhoven University it has been measured that for Al 2 O 3 layers with a thickness of 100 nm the increase of the oxygen plasma exposure time from 2 s to 10 s resulted in a lower carbon content, going from 2.5 ± 1.0 at.% for the standard ALD process, to < 1 at.% for the ALD process in which the oxygen plasma step in enlarged up to 10 s. In order to determine the inuence of this longer oxygen plasma exposure step, again the WVTR is determined and compared to WVTR of the samples of the previous section. This is both done for the extrinsic WVTR (Deltaperm measurements, sect. (3.1)) and the intrinsic WVTR (Ca test, (3.2)). From these results, more insight will be gained on the nucleation process of Al 2 O 3 layers on top of HDR SiO x, deposited by PE-CVD. First, measurements are performed on the extrinsic WVTR. The results are shown in g. (4.14). In this graph, the results of the samples with a 10 s oxygen plasma exposure time in each ALD cycle are compared to the samples which are deposited by the standard ALD step times. It can be seen that for thick layers of Al 2 O 3 (5 nm) the barrier of both series is about the same. The dierence in carbon content does not lead to a signicant improvement of the barrier quality. However, for thin sealing layers there is a nice improvement. Especially for the sample which was exposed to 11 ALD cycles ( 1.5 nm). For the standard deposition process this sample was at the critical thickness, where the WVTR dropped drastically, but for longer oxygen plasma exposure times, the WVTR already dropped to the low WVTR regime. With this deposition process, the 11 cycles of Al 2 O 3 ALD are already enough to seal the pores of the HDR SiO x, while with the standard deposition process it was not possible. The second method which is used to analyze the result of enlarging the oxygen plasma exposure time is a Ca test. In this test, the intrinsic WVTR of the samples is determined as well as that the number of pores of interconnecting defects can be determined. These defects are the main pathway of the permeation of water vapor. The results of that are shown in g. (4.15) and (4.16). Again these results are compared to the samples which were deposited at standard ALD conditions. From gures (4.14), (4.15) and (4.16) it can be concluded that for thin (1 1.5 nm) Al 2 O 3 layers there is a signicant improvement in the barrier performance. Both the extrinsic WVTR (Deltaperm) and the intrinsic WVTR (Ca test) are decreased, as well as the number of pathways allowing permeation (pinholes or pores of interconnecting defects). For the samples which were exposed to more ALD cycles (15 38 cycles = 2 5 nm), however, the WVTR values and the number of defects are the same within the error of the measurements. Since the inuence of a reduced carbon content would show up especially 54

57 Figure 4.14: Extrinsic WVTR, as determined by Deltaperm measurements. The results of the samples with the enlarged oxygen plasma exposure are shown in light blue (diamonds). As a reference the results of the standard samples are shown (dark blue, squares). Figure 4.15: Intrinsic WVTR, as determined by Ca test measurements. The results of the samples with the enlarged oxygen plasma exposure are shown in light blue (diamonds). As a reference the results of the standard samples are shown (dark blue, squares). 55

58 Figure 4.16: Total number of defects, as determined by Ca test measurements. The defects at the edge of the 5 5 mm 2 are excluded. The results of the samples with the enlarged oxygen plasma exposure are shown in light blue (diamonds). As a reference the results of the standard samples are shown (dark blue, squares). Also the results of the SiN x reference sample are shown. for thicker Al 2 O 3 layers, it can be concluded that the inuence of a reduced carbon content is none, or at least negligible, because for the thicker Al 2 O 3 layers in the deposited series, nice layer-by-layer growth leads to a crystalline structure in which the carbon atoms will cause defects. The increased exposure time of the surface to the oxygen plasma is only inuencing the thin Al 2 O 3 layers, indicating that there is an interaction between the surface and the oxygen plasma. The SiO x layer is normally terminated by OH groups at the surface, which would not react with the oxygen plasma. Therefore, and on the basis of the results presented and discussed in the previous alinea, it is proposed that the SiO x surface is contaminated with carbon. It can be assumed that carbon is contaminating the surface, because of the transport of the samples and the exposure to air in the dierent cleanrooms. This could be conrmed by (angle resolved) X-ray photo-emission spectroscopy. If there are carbon atoms attached to the surface, the sealing of the pores will be hindered, because the pores are not easily accessible for the growth of Al 2 O 3 anymore. Therefore it is still possible to form pathways of interconnecting defects via the defects in the Al 2 O 3 layer, the contaminating carbon and the pores in the SiO x layer. By the oxygen plasma exposure, the carbon can be removed to form CO and CO 2.[34] This is again a burning process, similar to the carbon which is removed from the TMA molecule in a PA-ALD step. When 56

59 the surface is exposed to the oxygen plasma for a longer time, more carbon can be removed, resulting in a better sealing of the pores. 4.5 Pore analysis via activation energy In the last section of this chapter on the results of the experiments - and the discussion about it - the activation energy of the barriers is determined. The activation energy can provide information on the pore structure of the barrier[6] as well as that it can be used to determine the barrier performance at ambient conditions - 20 ºC / 50 % relative humidity (RH) - from the results at elevated temperatures.[4] In order to further research the barrier structure, together with the determination of the barrier performance at ambient conditions, these activation energy experiments are performed, using a barrier of HDR SiO x, sealed with 2 nm of Al 2 O 3, deposited with a longer oxygen plasma exposure time in the oxidizing step of the ALD cycle, as extensively explained in the previous section. In the activation energy experiments, the logarithm of the WVTR values is plotted as a function of the inversed temperature. Such a plot is called an Arrhenius plot, leaving the slope of this plot linearly related to the activation energy, as shown in eq. (1.5). In the rst part of this section, the activation energy is determined, using WVTR measurements at 40, 50, 60 and 70 ºC, while in the second part the extrapolation to ambient conditions is done. In the rst part of this section, the activation energy is determined via the determination of the WVTR at 40 to 70 ºC. The results are shown in g. (4.17), using WVTR values of 1.83, 5.44, 13.2 and g/m 2 day at temperatures of 40, 50, 60 and 70 ºC. A linear t of these results (also shown) and the use of eq. (1.5), then leads to our activation energy, E act 76 kj /mole. However, it can be seen that for the higher temperature measurement points, the WVTR value (and their logarithm) is slightly lower than expected. Unexpected WVTR values at higher temperatures can be a result of the physical contact between the samples and the box they are placed in. The lms get deformed due to the high pressure dierence between the high pressure and low pressure side of the Deltaperm (up to 500 Torr). The apparent activation energy can also be used to determine the water vapor transmission rate at other conditions, e.g. at 20 ºC. This is done in g. (4.18). After subtraction of the aluminum permeation results, the activation energy could again be determined. Following this procedure, a higher activation energy is determined of E act 82 kj /mole. This can be used to extrapolate to 20 ºC and 90 % RH, resulting the WVTR to be W V T R = g/m 2 day. Using this results, it is also possible determine the water 57

60 Figure 4.17: Arrhenius plot, which can be used for the determination of E act. Using eq. 1.4 and eq. 1.5, the slope of the t can be used to determine the activation energy. vapor transmission rate at ambient conditions, being 20 ºC and 50 % RH. In that case, eq. (3.1), can be used resulting that W V T R = g/m 2 day. This WVTR value is an extrinsic value, determined via rigorous extrapolation, but which can be compared to other results, found in literature.[5, 25] The activation energy yields information on the pore structure of the barrier, because E act determines the energy which is needed to squeeze molecules of water vapor through the pores of the barrier.[33] The smaller the size of the pores, the higher this activation energy. For water vapor, this activation energy can vary from about kj /mol for a polymer substrate as PET, up to 84 kj /mol for glass like SiO 2 silica.[6, 27] The value of the activation energy ranges in this region, depending on the barrier quality.[6] It can be seen that the determined activation energy is a close to the activation energy of glass after subtraction of the background permeation; due to the sealing, only very small pores remain, through which water vapor almost cannot permeate. However, the value of the activation energy might be a bit too high by the degradation of the barrier, as explained further in this paragraph. The measurements for this activation energy are performed on the same sample; the sample is left in the Deltaperm while the conditions are changed, in order to continu with the next experiment. Therefore, it should be taken into account that the sample might be damaged, as mentioned earlier in this paragraph. Both due mechanical and physical detoriation, the measured WVTR for higher temperatures might be higher than if a fresh 58

61 Figure 4.18: Using E act after the subtraction of the aluminum foil tests, the WVTR can be extrapolated. This results in an activation energy of E act 82 kj /mole and a WVTR of g/m 2 day at 20 ºC and 90 % RH. sample would be measured. The possible damaging of the samples is mainly for the higher tempearture, these values are determined after the measurements at lower temperatures. Due to the high vapor pressure at these elevated temperatures the sample might bend, which can cause cracks in the barrier layer, as well as that the sample can physically touch the setup, damaging the very thin Al 2 O 3 layer (2 nm). On the other hand it is possible that the Al 2 O 3 layer is deteriorating by the presence of water vapor.[5] Via an exothermic reaction, Al 2 O 3 + 3H 2 O 2Al(OH) 3, a more amorphous structure might be formed, resulting in a higher WVTR. In the experiment by Dameron et al.[5] this deterioration of the Al 2 O 3 layer only shows up when this layer is directly exposed to water vapor, while no deterioration is seen when the backside of the barrier is exposed to water vapor. The latter situation is the same as for the results presented here. However, barrier performance of the polymer substrate and the HDR SiO x layer is bad and the Al 2 O 3 layer is very thin. Therefore, this eect cannot be fully excluded and should be researched further. Espescially if one keeps in mind that the presented barrier should protect devices from water vapour and when doing so, they are exposed to water vapor. Due to these to deteriation mechanisms, the barrier will perform worse at higher temperatures. The measured activation energy might therefore be too high. However, it still stands that the nal barrier (consisting of both HDR SiO x and Al 2 O 3 ) has only very 59

62 narrow pores of permeation, because pores of the HDR SiO x layer are sealed by the Al 2 O 3. 60

63 61

64 Chapter 5 Conclusions & Recommendations In this project it is shown that it is possible to increase the deposition rate of SiO x barrier layers, together with an increase in barrier performance. This is the result of the switch from LDR SiO x to HDR SiO x. When sealing this HDR SiO x with a thin layer of Al 2 O 3, the higher deposition rate can be accompanied by a low WVTR. With this technique, it is shown that only 15 ALD cycles of Al 2 O 3 ( 2 nm) are enough to decrease the WVTR down to g/m 2 day. This is an increase in barrier performance of a factor 350 compared to bare HDR SiO x on PEN and a factor 2-3 improvement compared to LDR SiO x (WVTR = g/m 2 day). The factor 350 improvement cannot be explained by ideal-laminate-theory.[27] The switch from LDR SiO x to HDR SiO x also leads to a large improvement in deposition rate. The SiO x layers can be deposited 8 times faster. The sealing with Al 2 O 3 ALD, however, is a relatively slow deposition technique, which can only be used because very thin layers of Al 2 O 3 are obtained. The current ALD deposition process is also too slow, but with the use of spatial ALD, problems with the growth speed of ALD (see eq. (2.11)) can be overcome.[24] The large increase in barrier performance by a thin lm of Al 2 O 3, deposited by ALD, can be explained by the sealing of the interconnected nano-pores of the HDR SiO x layer. This sealing is possible because the HDR SiO x consists of grains, while at the edges of these grains, nano-pores are present. Permeation of water vapor is allowed by the interconnection of those nano-pores. Blocking of these interconnected nano-pores, prevents water vapor from permeating through the barrier. Permeation is forced through the intrinsic material, which results in a drop in the WVTR. The drop in WVTR after a critical thickness, is observed after 2 nm for the extrinsic WVTR, but also for the intrinsic WVTR and the number of macro-defects and intercon- 62

65 nected nano-pores. It is also shown that the WVTR of the intrinsic material can reach g/m 2 day. Furthermore, this critical thickness decreases if the oxygen plasma exposure time in the ALD cycle is increased from 2.0 s to 10.0 s. On the other hand, when more ALD cycles are used, there is no change observed in the WVTR or the number of macro-defects and remaining interconnected nano-pores. This indicates that there is carbon present at the interface between the HDR SiO x buer layer and the Al 2 O 3 sealing layer. The oxygen plasma helps to remove the carbon, which improves the sealing of the interconnected nano-pores. The sealing of the pores is also shown by the activation energy of the deposited barriers. For standard SiO x barriers, deposited by CVD, E act ranges from kj /mol for poor barriers (depending on the substrate) up to 84 kj /mol for ideal glass-like SiO x barriers.[6] Although the barriers, as deposited in this research, are not purely SiO x barriers, their activation energy ( E act = 82 kj /mol) shows that the pores are sealed eciently. Because this E act is close to the ideal value of SiO x barriers, it can be concluded that the number of pathways through which diusion of water vapor is possible is decreased. At last, this activation energy is used to determined the WVTR at atmospheric conditions (20 ºC and 50 % RH). Using rigorous extrapolations of the WVTR values as measured by Deltaperm at elevated temperatures, it can be determined that the WVTR at atmospheric conditions is g/m 2 day. Outlook Further understanding of this sealing principle is important in its application for the deposition of barriers layers. Therefore, other measurements are proposed to proof and to understand the sealing of the pores in HDR SiO x. After the deposition of the sealing layer, Angle-Resolved X-Ray Photo-emission Spectroscopy (AR-XPS) measurements are proposed, so that the penetration of the Al 2 O 3 into the pores of the SiO x can be determined and compared to other buer layers (e.g. LDR SiO x or silicon wafer). Next to that, if the Al 2 O 3 is sealing the pores on the inside (as shown in g. (4.13)), for the rst ALD cycles the increase in mass of the samples should be larger. This mass increase could be measured in-situ by a quartz crystal microbalance (QCM). Various polymer substrates were already researched, using this method.[36] On the other hand, as discussed in sect. (4.5), also the degradation of the Al 2 O 3 sealing layers should be taken into account. It is suggested that they degrade relatively quickly.[5] This should also be researched before a production process of barrier layers can take place. This can be researched by exposing the barriers to water vapor for longer times 63

66 or at higher temperatures. Part of the increase in barrrier performance by a thin Al 2 O 3 layer might also explained by the fact that the SiO x layer works as a planarizing layer. When depositing Al 2 O 3 by ALD on polymer substrates, nucleation of the Al 2 O 3 can take place inside the substrate.[36] Therefore, more cycles are needed to obtain nice layer-by-layer growth with ALD. If the SiO x layer can work as a planarizing layer, none or much less nucleation can take place inside the layer, while the conformal, defect-free growth can start directly. In this way a thin Al 2 O 3 can work much better as a barrier layer. This eect will also play a role, although it is not really likely that just 2 nm of Al 2 O 3 works as such a good barrier. A better understanding of the sealing principle can lead to an increase in barrier performance. Next to that, it is shown that it is very important to keep the samples clean[4], especially between the deposition of the buer layer and the sealing layer. Particle contamination will lead to worse sealing, which decreases the barrier performance. Therefore extremely clean transport should be considered or both deposition steps should be performed roll-to-roll, so that particle contamination is reduced largely. For the carbon contamination by the exposure to air, as discussed is sect. (4.4), it should be researched if a cleaning process is possible, using an oxygen plasma, in order to remove all carbon and carbon containing materials. A better sealing of the pores becomes possible in this way, which increases the barrier performance. With this research, I hope that Fujilm can continue to increase the performance of their barrier layers, so that they can work as protection layers for exible electronics in the future. 64

67 65

68 Appendix A WVTR Measurement values after aluminum foil experiments It is explained in sec. (3.1) that the WVTR values, as measured by Deltaperm, are actually lower than the results shown in chap. (4). Experiments are performed using a 100 µm aluminum foil, to determine the permeation for an approximately impermeable foil. The remaining permeation is an estimate for the oset or leakage of the Deltaperm setup. In this appendix, the results are shown after subtraction of this oset. The oset is (3.6±0.5) 10 4 g/m 2 day. Unless it is mentioned otherwise, the measurements are performed at our standard conditions, 40 ºC and 90 % RH. The outline of this chapter will be the same as in chap. (4). This means the WVTR results of the single layer barriers will be discussed rst. After that, the WVTR results of the HDR SiO x barriers which are sealed with a thin layer of Al 2 O 3 ; rst in the standard ALD process and later with an enlarged oxygen plasma exposure time. The last part that will be discussed is the eect of this leakage on the determination o the activation energy. Single layer barrier results In sect. (4.2), three dierent barriers were discussed: LDR SiO x, HDR SiO x and Al 2 O 3. Each of them has an optimal thickness and an associated deposition rate. These properties are shown in g. (A.1), together with the original WVTR, as measured by Deltaperm, and WVTR corrected for leakage permeation. Only the results on PEN substrate are shown. It can be seen that for barriers with a large permeation the inuence of this leakage is negligable, while for really good barrier their is a real inuence. Barriers with a sealing layer Further in this research, bi-layer barriers are discussed. In this case the HDR SiO x is sealed with a thin layer of Al 2 O 3. This sealing process 66

69 Figure A.1: Original and corrected WVTR results of the single barrriers investigated in this research. Also their optimal thickness and associated deposition rate is shown. results in a barrier with good barrier properties, but which can also be deposited much faster than the standard LDR SiO x (a factor 8). The Al 2 O 3 sealing layer is very thin ( 5 nm), so that this layer can also be deposited relatively fast. ALD is a relatively slow deposition technique, but since the sealing is very thin, deposition can still be performed quickly. The sealing procedure of the HDR SiO x layer with Al 2 O 3 results in a barrier performance of g/m 2 day for a 5 nm sealing layer. When using a sealing layer of 20 nm, this can even go down to g/m 2 day. These results are shown in g. (A.2). The results are shown as a function of the number of ALD cycles. The growth rate per cycle is determined to be 0.13 nm /cycle. Using looxygen plasma exposure times in ALD cycle Also for the permeation experiments on the samples which are deposited with a longer oxygen plasma time step in the ALD cycle the WVTR can be corrected for small leakages is the Deltaperm system. The results of that are shown in g. A.3: for a 5 nm sealing layer, the WVTR is now improved to g/m 2 day. Also the improvement for the sample at the critical thickness of 1.5 nm is improved even further. Activation energy The last measurements which are inuenced by leakages, measured by testing the permeation of an impermeable aluminum foil, are the measurements for the determination of the activation energy. However, in order to make a correct extrapolation 67

70 Figure A.2: Al2O3 on HDR SiOx on PEN Figure A.3: longer plasma times 68